Skip to the Main Content

Note:These pages make extensive use of the latest XHTML and CSS Standards. They ought to look great in any standards-compliant modern browser. Unfortunately, they will probably look horrible in older browsers, like Netscape 4.x and IE 4.x. Moreover, many posts use MathML, which is, currently only supported in Mozilla. My best suggestion (and you will thank me when surfing an ever-increasing number of sites on the web which have been crafted to use the new standards) is to upgrade to the latest version of your browser. If that's not possible, consider moving to the Standards-compliant and open-source Mozilla browser.

November 1, 2006

Klein 2-Geometry VII

Posted by David Corfield

Let’s reconvene the latest session of the Honorable Guild of Categorifiers of Kleinian Geometry. I’ll briefly sum up what I learned from last month’s efforts. Our plan had been to work out the projective 2-space associated to a Baez-Crans (BC) 2-vector space, find the 2-group of projective linear transformations, and then study sub-2-groups, in order to throw up 2-geometries which were categorifications of sub-geometries of projective geometry such as Euclidean or spherical geometry. But Urs posed for us the task of finding the general linear 2-group of such 2-vector spaces, and he helped it see the light of day, we think, here.

This suggested another path to Euclidean 2-geometry if we could find a way to put an inner product on a BC 2-vector space, and then look at the sub-2-group of transformations which preserve it. However, we met with a small problem. I wondered whether we might look to other forms of 2-vector space, CC-modules for categories other than Disc(kk), such as what we called (1,1) vector 2-spaces.

Elsewhere, David Roberts wondered whether we could use 2-ordinals to keep track of incidence relations between the objects of our 2-geometries.

Tim Silverman joined the team and wrote many comments, perhaps he would like to sum up his discoveries.

Posted at November 1, 2006 9:19 AM UTC

TrackBack URL for this Entry:   http://golem.ph.utexas.edu/cgi-bin/MT-3.0/dxy-tb.fcgi/1014

49 Comments & 3 Trackbacks

categorified scalar product

put an inner product on a BC 2-vector space

I have changed my mind about the nicest way to put an inner product on the 2-space. It might be good to use categories internal to hermitean vector spaces instead of just plain vector spaces. Then a good notion of categorified inner product should be given by the internal hom\mathrm{hom}, making contact with the way I envisioned generalized Hilbert spaces of states here.

Posted by: urs on November 1, 2006 10:08 AM | Permalink | Reply to this

Re: categorified scalar product

categories internal to hermitean vector spaces

Can these be seen as CC-modules for some category CC?

Posted by: David Corfield on November 1, 2006 3:32 PM | Permalink | Reply to this

Re: categorified scalar product

categories internal to hermitean vector spaces

Can these be seen as C-modules for some category C?

I would say they are still Disc(k)\mathrm{Disc}(k)-modules - but with extra structure.

Like an ordinary hermitean kk-vector space is still a kk-module, but with additional structure.

Posted by: urs on November 1, 2006 6:08 PM | Permalink | Reply to this

Re: categorified scalar product

So what’s the sub-2-group of the 2-group you found (the Schreiber 2-group?) of linear transformations of a BC 2-vector space which preserve this additional structure? Something like, when k=k = \mathbb{C}, G=U(n)×U(m)G=U(n) \times U(m) worth of objects, with some subgroup of linear maps from k nk^n to k mk^m as morphisms between two objects, and the restricted action of the one you described?

Posted by: David Corfield on November 2, 2006 12:34 PM | Permalink | Reply to this

Re: categorified scalar product

[…] preserve this additional structure?

I am afraid, unfortunately, that I still haven’t said exactly what additional structure I want to regard as a scalar product on a BC 2-vector space.

But, as you indicate, once we agree on one definition, we should certainly compute the sub2-group of the general linear 2-transformation preserving this.

Posted by: urs on November 2, 2006 1:54 PM | Permalink | Reply to this

Re: categorified scalar product

According to your characterisation as categories internal to the category of hermitian vector spaces, I take it that the inner product on arrows applies to the whole vector space of arrows. If we stick with skeletal such 2-vector spaces, would you envisage there being non-zero inner products between arrows based at different objects?

Posted by: David Corfield on November 2, 2006 3:12 PM | Permalink | Reply to this

Re: categorified scalar product

According to your characterisation […]

The trouble is, I keep changing my mind about what a good notion of inner product would be. Sorry for that.

Right now I am trying to find one that generalizes the notion of the Killing form on a Lie algebra.

So consider for a moment a BC 2-vector space with the structure of a Lie 2-algebra on it. Call it

(1)L. L \,.

We get a representation of LL on itself: the adjoint 2-rep

(2)ad:LEnd(L). \mathrm{ad} : L \to \mathrm{End}(L) \,.

Bracketing by identity morphisms on objects xObj(L)x \in \mathrm{Obj}(L) yields linear functors L[x,]LL \stackrel{[x,\cdot]}{\to} L, and morphisms (xfy)Mor(L)(x \stackrel{f}{\to} y) \in \mathrm{Mor}(L) yield linear natural transformation between these

(3) ad(x) L ad(f) L ad(y). \array{ &\nearrow \searrow^{\mathrm{ad}(x)} \\ L & \Downarrow \mathrm{ad}(f)& L \\ & \searrow \nearrow_{\mathrm{ad}(y)} } \,.

The point is that it might be easier to see the right inner product on ad(L)\mathrm{ad}(L), because that should come from a categorification of the trace.

For monoidal 1-categories with duals on objects, there is a nice way to write the trace of any morphism by closing the string representing the morphism on itself.

Something similar should hold here, but right now I have no good grasp of it.

But as a warmup, we can compute “single matrix elements”, at least:

Consider Disc(k)\mathrm{Disc}(k) as a BC 2-vector space. We have

(4)LHom(Disc(k),L). L \simeq \mathrm{Hom}(\mathrm{Disc}(k),L) \,.

So yet another way to think of a morphism in LL is as a natural transformation

(5) x Disc(k) f L y. \array{ &\nearrow \searrow^{x} \\ \mathrm{Disc}(k) & \Downarrow f& L \\ & \searrow \nearrow_{y} } \,.

Similarly, a co-2-vector would be

(6) x * L f * Disc(k) y *. \array{ &\nearrow \searrow^{x^*} \\ L & \Downarrow f^*& \mathrm{Disc}(k) \\ & \searrow \nearrow_{y^*} } \,.

If we let {x if jx i+t(f j)}\{x_i \stackrel{f_j}{\to} x_i + t(f_j)\} be a basis for LL, and denote the dual basis by stars, then the (i,k)(i,k)-object-entry of the matrix representing

(7)ad(xfy) \mathrm{ad}(x \stackrel{f}{\to}y)

in that basis would be

(8)Disc(k)x iLad(x)Lx k *Disc(k). \mathrm{Disc}(k) \stackrel{x_i}{\to} L \stackrel{\mathrm{ad}(x)}{\to} L \stackrel{x_k^*}{\to} \mathrm{Disc}(k) \,.

Same as for 1-vector spaces.

But now we possibly also want to look at morphisms from and into Σ(k)\Sigma(k) to get matrix elements at the level of morphisms

(9) adx Σ(k) r L adf L r * Σ(k) ady . \array{ &\nearrow \searrow &&\nearrow \searrow^{\mathrm{ad}x} &&\nearrow \searrow \\ \Sigma(k) & \Downarrow r& L & \Downarrow \mathrm{ad}f& L & \Downarrow r^*& \Sigma(k) \\ & \searrow \nearrow && \searrow \nearrow_{\mathrm{ad}y} && \searrow \nearrow } \,.

Maybe now we can sum over rr to get a trace and hence a scalar product.

But I don’t know. Have to run now.

Posted by: urs on November 3, 2006 4:20 PM | Permalink | Reply to this

equivalence classes of n–vector spaces

I have asked the experts and found the following answer to a crucial question in this BC nn-vector business:

Question: We may regard an nn-term chain complex as living in an nn-category, whose morphisms are chain maps, 2-morphisms are chain homotopies, and so on.

Are equivalence classes in this nn-category in bijection with cohomology classes, i.e. with chain complexes all whose maps are the 0-map?

Answer: yes, this is true. It is in fact true for chain complexes in any category of projective modules.

This means that whenever one sees people inverting “quasi isomorphisms” in the 1-category of chain complexes, this is a 1-categorical workaround for talking about equivalence in a setup that is really ω\omega-categorical.

Hence forming the derived category of chain complexes is a lot like working with the full ω\omega-category structure on chain complexes in the first place (i.e. with reagrding arbitrary chain complexes as ω\omega-vector spaces).

I gather that this is very well known to the experts. But I didn’t know it before.

Posted by: urs on November 1, 2006 10:54 AM | Permalink | Reply to this

Re: Klein 2-Geometry VII

David said, looking pointedly at me,

Tim Silverman joined the team and wrote many comments, perhaps he would like to sum up his discoveries.

Er … [stands nervously at front of class]

[fiddles with papers]

[squeaky voice … ]

I think that blizzard of comments was more a sign of me getting myself oriented, spotting connections between things, and going a short way up a lot of paths, than producing technical results of any interest. Given that, I think I’ll summarise my overall impression of the directions we might go in, rather than saying too much about the ground I’ve covered so far.

One danger that we seem to have encountered is that, if we start from a beautifully symmetrical and simple Klein 1-geometry and try to categorify it, the result may be rather boring, in the sense that it won’t really introduce anything new. I think this showed up with the 2-vector spaces. It appears to me that JB and David all but cleared this out right back at the beginning of that discussion, and most of what was left was pretty much finished off by Urs.

Of course, my opinion about this may perhaps just be due to my own idiosyncratic view of what counts as boring and what as interesting …

In the course of trying to get round this, I explored a couple of different paths:

1) Looking at small, runty symmetry 2-groups instead of large, beautiful ones, or examining small, runty sub-2-groups of large beautiful 2-groups. I think (in retrospect) that this is more-or-less what I was doing in the series of posts where I looked at the 2-term chain complexes of 0-forms and 1-forms over various graphs. I think that the symmetries of the graphs induce unexpected and interesting sub-2-groups of the automorphism 2-group of the chain complex itself, along with some at least mildly interesting quotient spaces. I got a bit confused and bogged down with this, from posting too fast and thinking too slowly, but I think (or hope) there is still a lot of mileage in this approach and I want to come back to it when I have sorted out some of the sub-issues in my mind.

2) Avoiding going straight down to the skeletal 2-space. I seem to be a bit out of step with everybody else here. Of course, I don’t want to be evil (i.e. forget equivalences and only think about isomorphisms). At least, not very evil. On the other hand, I still think the process by which equivalences are carried out has the potential to be interesting in particular cases. It’s not like the difference between R nR^n and a single point is of no interest whatever …

From the particularities of 2-vector spaces, I then jumped up to the general case of beautifully symmetrical 2-spaces, in particular taking 1-spaces with a transitive group action on them, and constructing internal categories in categories of these spaces. While this symmetry requirement to some extent discards the interest of going to 2-spaces (e.g. it avoids going in the orbifold/stack direction of keeping track of different kinds of points), it at least lets us get an overview of all those boringly symmetrical 2-spaces.

This ‘internal category’ approach more-or-less forces us to think about the 2-space as a groupoid, and to think about the relationship between this groupoid, and the weak quotient groupoid of the object 1-space by its automorphism group (qua 1-space).

The groupoid view places at least two phenomena in front of us: isomorphisms between different points, and automorphisms of single points. It is probably a good idea to look at these separately, at least initially, although they have a good deal in common.

I tried approaching both of these through a fibre-bundle view. We can, perhaps, think of the automorphisms of a point as forming a fibre over that point; and of the objects and morphisms in one component of a groupoid as forming a kind of categorified fibre over the component (i.e. thinking of a base space whose points are the components, with the fibres being the internal structure of the components). The most obvious odd thing (to me) about this manner of thinking is that the paths in the space are global automorphisms of the base space, rather than, e.g. smooth curves. This is of course usual for Klein geometry.

I think this bundle approach, and related approaches, are worth exploring in more depth, because they may quickly connect up to a lot of other things. Judging by other threads, they’re also, perhaps, the approaches most likely to be of interest to other participants here. But I’m kind of reluctant to do too much on this, because all those other participants know such an awful lot more about this sort of thing than I do … so anything I say is liable to sound rather babyish.

At the moment, I’m having a go at working on the problem from the opposite direction: instead of picking a 2-space and trying to understand its 2-symmetry group, its figures, and its figure-symmetry sub-2-groups, I’m trying to pick a particular, very simple, 2-group and look at its various actions, as well as generally trying to understand the structure of the 2-group as thoroughly as possible.

There seem to be a very large number of different generalisations of the concepts of ‘action’ and ‘quotient’. I want to understand these and classify them. Once I’ve done that, I think a bunch of other things will become clearer.

One other point I’d like to make is that ordinary group theory is a huge, complicated subject in its own right. It’s not surprising if the theory of 2-groups has a lot of stuff lurking below the surface which isn’t simply a straightforward generalisation of the 1-group case. There’s really a tremendous lot to discover here. I really ought to familiarise myself with the existing literature, but I’m not sure I’m even at the point where I can understand it, except maybe at a very superficial and useless level.

I very much like Dave Roberts’ idea of attacking generalisations of the subobject lattice, and looking at 2-dimension via 2-ordinals. On the other hand, I’m increasingly of the opinion that the best way to categorify incidence geometry is to creep up on it from behind, rather than coming at it head-on.

To summarise the summary:

2-symmetry 2-groups which are categorifications of symmetry groups seem to have interesting 2-subgroups which are not simply categorifications of the subgroups of the symmetry groups.

Not all spaces are skeletal, and equivalences have interest beyond their mere existence!

There are some interesting things to be said about spaces which are groupoids in two different ways.

It may be interesting to think about groupoids as fibre or 2-fibre bundles.

It may be worth picking a 2-group and understanding it in detail, looking at the ways it can act as a 2-symmetry 2-group. This should involve understanding its sub-2-group lattice, and getting a better handle on what a quotient is.

Generally, we can expect interesting 2-groups to contain a lot of stuff that goes beyond what we can understand in terms of 1-groups.

To summarise the summary of the summary

Graphs, groupoids, bundles, quotients … ohhh, it’s all hard!

Posted by: Tim Silverman on November 2, 2006 8:19 PM | Permalink | Reply to this

Re: Klein 2-Geometry VII

Urs has a nice clear formulation here of the action of morphisms on 2-morphisms in the general linear 2-group of transformations on a BC 2-vector space. Perhaps we might consider possible sub-2-groups of this 2-group. Presumably we can form those which are full on 2-morphisms but which restrict independently on the two components of the 1-morphisms, e.g., to orthogonal groups.

On the other hand, if we go for non-full on 2-morphisms options, then we’ll have to be careful that the action of the 1-morphisms preserves them.

Posted by: David Corfield on November 8, 2006 2:20 PM | Permalink | Reply to this

Re: Klein 2-Geometry VII

Back here:

Might it have appeared in HDA5?

I don’t recall having seen it there. What does appear is the slightly similar Poincaré 2-group.

Urs commented on the similarity between his 2-group and the Poincaré 2-group.

I guess actually it’s just a full sub-2-group of his 2-group for p=4p=4 and q=1q=1, with the product of the trivial subgroup of GL(1) and the Lorentz subgroup of GL(4) for 1-morphisms.

This perspective opens up a huge range of 2-groups. We might also look at some arithmetic ones, with 1-morphisms some subgroup of SL(n,Z)×SL(m,Z)SL(n,Z) \times SL(m,Z), and 2-morphisms maps between Z nZ^n and Z mZ^m. When do we start on the 2-Langlands Program?

Then we might try permutation groups, although perhaps these are just going to be sub-2-groups of Urs’ 2-group for the ‘field with one element’ (see here, TWF184 and TWF187, and Tuesday 8 entry here).

Posted by: David Corfield on November 9, 2006 9:02 AM | Permalink | Reply to this

Re: Klein 2-Geometry VII

I guess actually it’s just a full sub-2-group of his 2-group for p=4p=4 and q=1q=1 […]

Oh, good point! Yes, that seems to be right. Interesting.

(The group of morphisms, strictly speaking, is Hom(k 4,k 1)\mathrm{Hom}(k^4,k^1), hence covectors of k 4k^4 instead of vectors, but of course that’s completely immaterial.)

When do we start on the 2-Langlands Program?

Uh, er - as soon as I find the time I’ll join you on that project. Meanwhile we have to stick to Kapranov’s observation on categorified Langlands.

Posted by: urs on November 9, 2006 9:48 AM | Permalink | Reply to this

Re: Klein 2-Geometry VII

I see (ref 23) John’s been working on the representation theory of the Poincaré 2-group. Would our discovery that it’s a sub-2-group help? Perhaps there’s a 2-adjunction between 2-categories of 2-representations, corresponding to the adjunction between induced and restricted representations at the level below.

Posted by: David Corfield on November 9, 2006 1:13 PM | Permalink | Reply to this

Re: Klein 2-Geometry VII

representation theory of the Poincaré 2-group

I am looking forward to seeing that.

I am aware of the Crane-Sheppeard work on reps of the Poincaré 2-group #.

It seems that in this context people always only consider 2-reps on Kapranov-Voevodsky 2-vector spaces (or a certain infinite-dimensional generalizations of them).

As I apparently don’t get tired of mentioning, it strikes me that we should really be looking, much more generally, at reps in bimodules. That’s because we have the chain of inclusions

(1)KV2VectBim VectMod \mathrm{KV}2\mathrm{Vect} \stackrel{\subset}{\to} \mathrm{Bim} \stackrel{\subset}{\to} {}_{\mathrm{Vect}}\mathrm{Mod}

and because reps in Bim\mathrm{Bim} show up at various places in string physics. These reps yield associated bundle gerbes, associated nonabelian bundle gerbes and string-bundles with connection, for instance.

And what is nice, for every strict 2-group

(2)G 2=(HG) G_2 = (H \to G)

we get a canonical 2-rep

(3)ρ:Σ(G 2)Bim \rho : \Sigma(G_2) \to \mathrm{Bim}

which works like

(4)ρ: g h g [H] g [H] h [H] [H] g, \rho \;\;:\;\; \array{ & \nearrow \searrow^{g} \\ \bullet &h \Downarrow \;& \bullet \\ & \searrow \nearrow^{g'} } \;\; \mapsto \;\; \array{ & \nearrow \searrow^{\mathbb{C}[H]_g} \\ \mathbb{C}[H] &\cdot h \Downarrow \;\;& \mathbb{C}[H] \\ & \searrow \nearrow^{\mathbb{C}[H]_{g'}} } \,,

where [H]\mathbb{C}[H] is the group algebra of HH, and [H] g\mathbb{C}[H]_g is this algebra regarded as a bimodule over itself, with the right action twisted by α(g)Aut(H)\alpha(g) \in \mathrm{Aut}(H).

I have no proof, but several indications that we should really be looking at 2-reps in bimodules this way. Apart from the fact that I know a couple of places where these actually appear prominently in applications, the reason is that

a) Bim\mathrm{Bim} is much larger than KV2Vect\mathrm{KV}2\mathrm{Vect}. And for instance the work by Baas/Dundas/Rognes/Richter indicates that 2-vector bundles with fibers in KV2Vect\mathrm{KV}2\mathrm{Vect} are not general enough, while the work by Stolz/Teichner indicates that those with fibers in Bim\mathrm{Bim} are.

b) several people indicated to me that the inclusion

(5)Bim VectMod \mathrm{Bim} \stackrel{\subset}{\to} {}_{\mathrm{Vect}}\mathrm{Mod}

should in fact be an equivalence

(6)Bim VectMod \mathrm{Bim} \simeq {}_{\mathrm{Vect}}\mathrm{Mod}

if conditions are chosen suitably.

So this then would mean that 2-reps in Bim\mathrm{Bim} capture actually all possible Vect\mathrm{Vect}-linear 2-reps.

But I am still looking for an actual proven statement of this supposed equivalence. If anyone out there knows about this, please drop me a note.

Would our discovery that it’s a sub-2-group help?

Hm, maybe. I’d need to think about that.

Posted by: urs on November 9, 2006 2:31 PM | Permalink | Reply to this

Re: Klein 2-Geometry VII

For some strange reason I wrote:

I am still looking for an actual proven statement

Actually, I do know the answer.

Posted by: urs on November 9, 2006 8:56 PM | Permalink | Reply to this

Re: Klein 2-Geometry VII

I wrote:

Actually, I do know the answer.

So let me get this straight, finally.

Viktor Ostrik’s theorem 1 (p.10) says that for CC a sufficiently well-behaved abelian monoidal category, the canonical monomorphism

(1)Bim(C) CMod \mathrm{Bim}(C) \to {}_C \mathrm{Mod}

is essentially surjective on indecomposable objects, i.e. surjective on equivalence classes of indecomposable objects.

Here “sufficiently well behaved” means CC is semisimple, rigid, has finitely many irreducible objects (but see remark 9) and has irreducible tensor unit.

A module category is indecomposable if it is not the direct sum of two nontrivial module categories.

Examples of categories CC with these properties are given on p. 5. Finite dimensional vector spaces are among them.

So:

we have a sequence of monomorphisms

(2)KV2VectBim(Vect) VectMod \mathrm{KV}2\mathrm{Vect} \stackrel{\subset}{\to} \mathrm{Bim}(\mathrm{Vect}) \stackrel{\subset}{\to} {}_\mathrm{Vect}\mathrm{Mod}

with the last one being essentially surjective on indecomposable objects.

Posted by: urs on November 10, 2006 5:19 PM | Permalink | Reply to this

Re: Klein 2-Geometry VII

Does the reason for the physical interest in the Poincaré 2-group relate to the interest in the Poincaré group? If so, wouldn’t the double cover of the 2-group be worth looking at?

I wonder how general we can make our family of skeletal 2-groups. Presumably any pair of groups GG and HH with a left and right action, respectively, on an abelian group of 2-morphisms, KK would do. Is there a term for what is to a representation as a bimodule is to a module? Birepresentation?

Posted by: David Corfield on November 10, 2006 3:00 PM | Permalink | Reply to this

Re: Klein 2-Geometry VII

Would it be useful to combine an internal symmetry group with the Poincaré 2-group? Something like 1-morphisms are G×HG \times H, where GG is SO(3,1)SO(3,1) (or its double cover) and HH is SU(3)×SU(2)×U(1)SU(3) \times SU(2) \times U(1) (divided by Z 6Z_6 if you like). Then for 2-morphisms one would need at least R 4R^4, and one could keep with that allowing HH to act trivially. On the other hand, there is scope to allow a larger space of 2-morphisms and a nontrivial action by HH.

Posted by: David Corfield on November 11, 2006 10:44 AM | Permalink | Reply to this

Re: Klein 2-Geometry VII

For what it’s worth, a paper on representations of the Poincaré 2-group and state sum models should appear in a month or so, by Laurent Freidel, Aristide Baratin, and some other authors possibly including myself. Here the Poincaré 2-group is represented on 2-Hilbert spaces of the Crane-Yetter sort, following the ideas of Crane and Sheppeard.

Posted by: John Baez on November 13, 2006 7:39 AM | Permalink | Reply to this

Re: Klein 2-Geometry VII

There must be a name for what I’m calling a birepresentation. It’s like a profunctor except taking values in Vect.

G×H opVectG \times H^{op} \rightarrow Vect

I was considering groups GG and HH (as one object categories), but any categories would do. Either way, could we then compose these bireps using Kan extensions as is done for profunctors here? If that’s right that every birep of groups gives a (strict) 2-group, then we’d have a new type of composition.

Posted by: David Corfield on November 13, 2006 9:04 AM | Permalink | Reply to this

Re: Klein 2-Geometry VII

Probably a trivial exercise with enriched coends, if I knew anything about them. For a similar use, but over a different category, see pp. 13 -14 of Lawvere’s Taking Categories Seriously.

I like the “Rejecting the complacent description of that [aforementioned] identification as a mere analogy or amusement, its relentless pursuit is continued” in the abstract. Here’s to the relentless pursuit of category theoretic ideas into concrete situations - Klein 2-geometry is no mere analogy or amusement either.

Posted by: David Corfield on November 13, 2006 6:13 PM | Permalink | Reply to this

Re: Klein 2-Geometry VII

David observed that the basic structure of what we seem to have identified as the general linear 2-group (on a skeletal 2-vector space) apparently generalizes to a situation where

the group of objects is a product G 1×G 2G_1 \times G_2

the group of morphisms is the additve group of a linear vector space VV

the target map is trivial

objects act on morphisms by means of a birep

G 1 op×G 2VectG_1^\mathrm{op} \times G_2 \to \mathrm{Vect}.

Inspired by this he observes:

Either way, could we then compose these bireps using Kan extensions as is done for profunctors here? If that’s right that every birep of groups gives a (strict) 2-group, then we’d have a new type of composition.

I see how the appearance of bireps (or profunctors, if one wishes) anywhere makes one want to define their composition - which is only natural.

But for the present context I seem to be at a loss with respect to understanding what this composition would mean abstractly for a 2-group.

And in fact, it seems that only on the morphisms a natural composition is induced. Do we really get a map

first 2-group times second 2-group to third 2-group

this way, that respects any of the structure around?

Posted by: urs on November 13, 2006 6:44 PM | Permalink | Reply to this

Re: Klein 2-Geometry VII

So you’re agreeing that the coend is possible, but don’t know whether it has a meaning? Or is there something about Vect that doesn’t allow enriched coends? Mimicking the quotienting process for Set, one would expect something like W×V/~W \times V /~ where (w,v)~(w,v)(w',v') ~ (w,v) if there’s an element of the middle group which acts on ww on the left to give ww' and on vv' on the right to give vv. Oh, where VV and WW are the vector spaces involved in the bireps.

I’ll see if it has a meaning.

Posted by: David Corfield on November 13, 2006 7:14 PM | Permalink | Reply to this

Re: Klein 2-Geometry VII

So you’re agreeing that the coend is possible, but don’t know whether it has a meaning?

Yes. I haven’t thought about it, but I would expect the composition of bireps you have in mind exists.

But I am puzzled by what that should mean for the corresponding 2-groups.

Posted by: urs on November 13, 2006 7:38 PM | Permalink | Reply to this

Re: Klein 2-Geometry VII

No meaning yet, but I kinda convinced myself that if we’re composing (G,H)(G,H) with birep VV and (H,K)(H,K) with birep WW, then the action of (G,K)(G,K) on that quotient I mentioned is the sum over h of (hwk 1,gvh 1)(hwk^{-1},gvh^{-1}). Seems to be invariant over the quotient class.

I’m casually slinging round the term ‘birep’, but does it exist? Maybe, there’s no point to it as a group is iso to its opposite and so a birep is just the same as a rep of the product. Unless, of course, we can compose suitable bireps, like bimodules.

Posted by: David Corfield on November 13, 2006 7:45 PM | Permalink | Reply to this

Re: Klein 2-Geometry VII

I’ll just say some things I already said in my recent long post, which may have been buried amid technicalities.

David wrote:

I’m casually slinging round the term ‘birep’, but does it exist?

I’ve never heard of it, so if you want to sound like everyone else, you can say “bimodule” instead.

Maybe, there’s no point to it as a group is iso to its opposite and so a birep is just the same as a rep of the product.

Right, there’s not much need to talk about birepresentations, since they’re just representations of a product of groups.

But, for the same reason, there isn’t much need to talk about bimodules of rings or algebras, and people do about those.

The advantage of having special words like this is that bimodules and birepresentations can be thought of as morphisms going from one ring, or algebra, or group, to another.

Unless, of course, we can compose suitable bireps, like bimodules.

You can indeed, since a birepresentation is a bimodule!

A representation of a group GG is the same as a module of its group algebra k[G]k[G], where kk is your favorite field.

A G,HG,H-birepresentation is thus the same as a k[G],k[H]k[G], k[H]-bimodule, or in other words a module of k[G]k[H]k[G] \otimes k[H], or in other words a module of k[G×H]k[G \times H].

So, composing birepresentations is just a special case of composing bimodules! If you hadn’t already defined how to compose birepresentations, this is how you should define it.

All this stuff is very, very related to Hecke algebras and Hecke operators - as I’ll explain someday, when I get around to saying what Jim has done over the last 2 years.

Posted by: John Baez on November 13, 2006 10:54 PM | Permalink | Reply to this

Re: Klein 2-Geometry VII

All this stuff is very, very related to Hecke algebras and Hecke operators - as I’ll explain someday, when I get around to saying what Jim has done over the last 2 years.

That’s reassuring. You’ve mentioned that work before. Do you realise the anxiety you cause by continually making additions to the list of things that Jim has worked on which may never see the light of day?

Posted by: David Corfield on November 14, 2006 8:48 AM | Permalink | Reply to this

Re: Klein 2-Geometry VII

The earlier mention was from Klein 2-geometry II:

In particular, the subject of GG-invariant binary relations underlies the theory of Hecke operators and Hecke algebras. This is a little appreciated fact, because most people working on Hecke algebras focus on very special groups GG - namely, finite reflection groups or SL(2,Z)SL(2,Z). Jim has been working out the general theory, and someday I’ll have to explain a bit of it on This Week’s Finds.

Perhaps I strayed close with my suggested 2-group here - 1-morphisms some subgroup of SL(n,Z)×SL(m,Z)SL(n,Z)\times SL(m,Z), and 2-morphisms maps between Z nZ^{n} and Z mZ^m.

Posted by: David Corfield on November 14, 2006 10:03 AM | Permalink | Reply to this

Re: Klein 2-Geometry VII

Do you realise the anxiety you cause by continually making additions to the list of things that Jim has worked on which may never see the light of day?

Fortunately, Jim gives talks about this stuff at UCR sometimes. So there is a fair chance that any particular idea will be explained in somebody’s notes. Nevertheless, it’s clear to me that his notebooks will keep researchers (after his death) busy for many years, full of unpublished but important and prescient ideas, like those that we find now in the notebooks of Leibniz. (OK, Jim probably won’t be as significant as Leibniz.)

Posted by: Toby Bartels on November 16, 2006 6:30 PM | Permalink | Reply to this

Re: Klein 2-Geometry VII

Back to the question of whether composition of bireps has any meaning for the corresponding 2-groups. Let’s consider the full general linear 2-groups for BC 2-vector spaces k p,qk^{p,q} and k q,rk^{q,r}. The former has as 2-morphisms functions between k pk^{p} and k qk^{q}, and similarly for the latter. Now the tensor product is formed of pairs of 2-morphisms (f,g)(f,g), but where (f.a,g)(f.a,g) is identified with (f,a.g)(f,a.g) for aa in GL(q)GL(q). So these 2-morphisms are maps from k pk^{p} to k rk^{r} which factor through k qk^{q}, and we don’t care how they factor.

Don’t you ever do something like that with bundles, where a fibre for one bundle is the base for another bundle?

Posted by: David Corfield on November 14, 2006 9:27 AM | Permalink | Reply to this

Re: Klein 2-Geometry VII

I should have said:

Don’t you ever do something like that with bundles, where a total space for one bundle is the base space for another bundle?

Hmm, so is there a category of topological spaces with maps continuous functions, f:XYf: X \to Y such that ff is a vector bundle, and the obvious composition. Then we’re looking at how the corresponding 2-groups of gauge transformations compose.

Posted by: David Corfield on November 14, 2006 11:32 AM | Permalink | Reply to this

Re: Klein 2-Geometry VII

Don’t you ever do something like that with bundles, where a total space for one bundle is the base space for another bundle?

Yes, we do. But could you explain again in which sense this is related to the composition of 2-groups that you have in mind? I am not sure yet that I see what you mean.

Posted by: urs on November 16, 2006 8:01 PM | Permalink | Reply to this

Re: Klein 2-Geometry VII

The intuition, which may of course be wrong, is: We can compose two BC 2-vector spaces, k p,qk^{p,q} and k q,rk^{q,r} to form a k p,rk^{p,r} space, noting the composition is skeletal. Then the composition of 2-groups I’m describing corresponds to the composition of the general linear 2-groups associated to the 2-vector spaces.

Posted by: David Corfield on November 17, 2006 9:09 AM | Permalink | Reply to this

Re: Klein 2-Geometry VII

I had been thinking to myself that this composition of 2-groups feels like composition of species (aka structure types), see here. And of course the latter is also just another example of a coend construction. There’s a paper by Fiore, Gambino, Hyland and Winksel which treats generalised species. Our ordinary ones are (1,1) species, a certain kind of profunctor. Composing species is then a form of composing profunctors (a bit more subtle than the usual composition, see p.10 onwards).

On another point, the authors say:

More recently, Baez and Dolan considered further generalisations of these structures leading to the concepts of sorted symmetric set-operad [1] and of stuff types [2] (see also [39]). The former, though not the latter, can be directly recast in our setting (p. 2)

So, I wonder how stuff types and composition of stuff types (p. 31) fits into the picture.

Then where is the theory of generalised stuff types?

If an operad is a monoid object in the category of structure types with composition as the monoidal structure, how about in the category of generalized structure types (species)? And in the category of stuff types?

Elsewhere John spoke of the search for gold nuggets. Sometimes they seem to be so liberally sprinkled around in the rock that it’s hard to know which to pick out first. Admittedly picking them out may not be as simple as it looks.

Posted by: David Corfield on November 17, 2006 4:58 PM | Permalink | Reply to this

Re: Klein 2-Geometry VII

David writes:

[…] is there something about Vect that doesn’t allow enriched coends?

No, it’s a cocomplete symmetric monoidal category so you’re fine.

(I’m assuming that you’re allowing infinite-dimensional vector spaces here.)

Posted by: John Baez on November 13, 2006 9:55 PM | Permalink | Reply to this

Re: Klein 2-Geometry VII

Okay, I’ll do a little work today. David writes:

There must be a name for what I’m calling a birepresentation. It’s like a profunctor except taking values in Vect. G×H opVect G \times H^{op} \to \mathrm{Vect}

I assume GG and HH here are groups, or maybe groupoids, or maybe categories?

If they’re groups, ordinary mortals would call this thing a “representation of G×HG \times H”, since H opHH^{op} \cong H for groups.

If they’re categories, I might call this thing a “Vect\mathrm{Vect}-enriched profunctor”. Australians would call it a “bimodule” or simply a “module”.

That last fact deserves some explanation. This spring in Chicago, Steve Lack gave a bunch of talks on Australian nn-category theory. The notes will eventually show up in the proceedings of the IMA conference. He explained that these days, Aussies are very fond of doing everything with profunctors instead of functors. And, ever since Kelly, they’ve been fond of doing everything in an enriched context whenever possible. So, anything you or I can do with a functor F:CD F : C \to D they can do with a KK-enriched profunctor F:C op×DK F: C^{op} \times D \to K where KK is a sufficiently nice category, like Vect\mathrm{Vect}.

This conceptual move is a lot like replacing functions from SS to TT by S×TS \times T-shaped matrices - going from set theory to linear algebra. Any function gives a matrix of 0’s and 1’s, so one doesn’t lose much - and one gains a lot.

But, they get sick of saying “K\mathrm{K}-enriched profunctor”, so they say “bimodule” or simply “module”.

Note that in ordinary ring theory, an A,BA,B-bimodule is just a module over A opBA^{op} \otimes B. So, there’s no real need for the word “bimodule” in that context, if one is feeling hasty. The same thing is happening here! In fact, if one is clever enough, one can see that what we’re doing subsumes the ordinary theory of bimodules of rings. (There’s a little wrinkle here which you’ll see if you try to actually to do this, but it’s no big deal.)

Anyway, in ordinary ring theory we get a bicategory of

  • rings
  • bimodules
  • bimodule homomorphisms

So, here we get a bicategory of

  • categories
  • KK-enriched profunctors
  • natural transformations between KK-enriched profunctors

and everything works more or less the same if KK is sufficiently nice - which it is in your example.

I don’t think you need to formally know anything about “KK-enriched coends” to deal with this stuff at an intuitive level. When you compose KK-enriched profunctors, it’s just a categorified version of matrix multiplication! The formula for matrix multiplication involves a sum of products. Similarly, the formula for composing KK-enriched profunctors involves a “weighted colimit of tensor products”. This sort of thing is also called a “KK-enriched coend”, but don’t let that perturb you. The main thing is, it works whenever KK is sufficiently nice - and Vect\mathrm{Vect} is sufficiently nice.

But, just for my own peace of mind, let me explain that wrinkle I mentioned. A ring is not just a category with one object, it’s an Ab\mathrm{Ab}-enriched category with one object. So, a bimodule of two rings AA and BB is not really a functor F:A op×BAb;F : A^{op} \times B \to \mathrm{Ab}; it’s an Ab\mathrm{Ab}-enriched functor F:A opBAb.F : A^{op} \otimes B \to \mathrm{Ab}.

So, to make our general “bimodules” really subsume the usual concept of bimodule for rings, it’s best if we let them be, not functors F:C op×DK F : C^{op} \times D \to K where CC and DD are categories, but KK-enriched functors F:C opDK F : C^{op} \otimes D \to K where CC and DD are KK-enriched categories!

Then everything in sight is KK-enriched, which makes life simpler in the end - we’re not hopping around between different contexts.

But, if your CC and DD are just ordinary categories, fear not! You can make them KK-enriched using any monoidal functor from Set\mathrm{Set} to KK.

For example, if K=VectK = \mathrm{Vect}, you can use the “free vector space on a set” functor. This amounts to taking your CC and DD and throwing in new morphisms that are linear combinations of the existing ones. We get some new Vect\mathrm{Vect}-enriched categories C˜\tilde{C} and D˜\tilde{D}. Then any plain old functor F:C op×DVect F: C^{op} \times D \to \mathrm{Vect} becomes a Vect\mathrm{Vect}-enriched functor F˜:C˜ opD˜Vect \tilde{F}: \tilde{C}^{op} \otimes \tilde{D} \to \mathrm{Vect} so we’re working in a completely enriched world.

If you think what this means when CC and DD are groups, you’ll see that C˜\tilde{C} and D˜\tilde{D} are their group algebras! A representation of a group is the same as a representation of its group algebra, so nothing is lost; we’re just more firmly planted in the world of linear algebra now.

So, in fact, what the Australians really do is generalize the bicategory of

  • rings
  • bimodules
  • bimodule homomorphisms

to the bicategory of

  • KK-enriched categories
  • KK-enriched profunctors
  • KK-enriched natural transformations between KK-enriched profunctors

This is what they use whenever wimps like ourselves might be tempted to use

  • categories
  • functors
  • natural transformations

I’m exaggerating somewhat, but this is roughly their attitude. Remember, this is the country where Crocodile Dundee came from.

Posted by: John Baez on November 13, 2006 9:52 PM | Permalink | Reply to this

Re: Klein 2-Geometry VII

Am I alone in wishing that the Australians would spend time exploiting this powerful apparatus to work out particular cases? Perhaps they are. If you read Lawvere’s article, mentioned here, you’ll see what can be achieved by ‘relentless pursuit’ even with a simple enriching VV such as the extended reals (p. 13 onwards). This makes clear that it’s not simply a case of automatically applying a general theory.

Posted by: David Corfield on November 14, 2006 8:58 AM | Permalink | Reply to this

Re: Klein 2-Geometry VII

David wrote:

Am I alone in wishing that the Australians would spend time exploiting this powerful apparatus to work out particular cases?

I’d rather let the Australians do what they do best, leaving the rest to us.

When one hears of “Australian category theorists”, one imagines massed ranks of powerful mathematicians. That indeed is the image the phrase is supposed to evoke. But, you can count them on two hands - and most of them are Ross Street.

When Street retires, I believe there will be no full professor at a research university in Australia who does category theory full time. Max Kelly is retired. I believe Steve Lack teaches at a school without mathematics grad students. Michael Johnson spends most of his energy on computer science. Brian Day has a loose affiliation with Macquarie, but not a paying job. Wesley Phoa is in finance; Dominic Verity was out of academia until recently - luckily he’s back in, and quite energetic. Alexei Davydov and the redoubtable Michael Batanin are research fellows. Do you know many more?

When Street retires, will anyone have sufficient clout to hire more people in this area? I hope so.

Luckily there are expatriate Australian category theorists like John Power and Mark Weber, honorary Australian category theorists like Martin Hyland - and would-be Australian category theorists like the people who attended Steve Lack’s talks at Chicago this spring, including Eugenia Cheng, Nick Gurski, Mike Shulman and me.

Posted by: John Baez on November 23, 2006 9:33 AM | Permalink | Reply to this

Re: Klein 2-Geometry VII

In that case.

It’s just that the wheels of n-category theory turn rather frustratingly slowly. Now we have you down on record saying that mathematical objects can be “intrinsically interesting”, you presumably also believe that the man hours devoted globally to mathematical research are not distributed fairly according to this intrinsic interest. A few more concrete successes might help redress the balance.

Posted by: David Corfield on November 23, 2006 3:51 PM | Permalink | Reply to this

Re: Klein 2-Geometry VII

David Corfield:

… the wheels of n-category theory turn rather frustratingly slowly.

You have a curiously impatient attitude towards mathematics.

Maybe I was like that too before I had lots of classes, half-written papers and books, grad students, grant proposals, committee work, letters of recommendation to write, conferences to organize, talks to plan for, travel to arrange - and blogging, shopping, socializing, gardening, eating, sleeping, etcetera to attend to. As a lonely youth, I had endless hours to spend on math. But now, getting to the point of doing math seems like pushing a large boulder encrusted with moss. Having ideas is quick. Proving theorems take a bit longer. But it’s all the other stuff that eats up vast amounts of time.

So, I’m learning to be patient.

And so, when you imagine “the Australian category theorists”, you should imagine Ross Street endlessly writing grant proposals and reports like this, this and this. That’s the reality.

Luckily, it’s not as if there’s really any rush to get everything figured out! I actually enjoy subjects when they’re mysterious more than when the solutions are all worked out. When I’m done unravelling a mystery, the fun is over and it’s time to move on.

… you presumably also believe that the man hours devoted globally to mathematical research are not distributed fairly according to this intrinsic interest.

No, nothing is perfect. In fact, most of the man hours supposedly devoted to mathematical research are actually spent doing chores of various sorts that prevent one from actually thinking. But that’s life!

Posted by: John Baez on November 23, 2006 8:45 PM | Permalink | Reply to this

Re: Klein 2-Geometry VII

You have a curiously impatient attitude towards mathematics.

Perhaps it’s just projected from my frustration with philosophy. As I mentioned in my post about my last conference:

Brendan Larvor, author of a very good book on Lakatos, opened his talk with “Here we are nearly 45 years after Lakatos’s Proofs and Refutations, still looking for the new epistemology…”. This struck a chord as Chapter 7 of my book opens “Nearly forty years have passed since Imre Lakatos published his paper Proofs and Refutations…”, itself updated from its earlier appearance as a journal article in 1997. We imagined ourselves 15 years hence, “Nearly 60 years have passed…”.

As you say,

Luckily, it’s not as if there’s really any rush to get everything figured out! I actually enjoy subjects when they’re mysterious more than when the solutions are all worked out. When I’m done unravelling a mystery, the fun is over and it’s time to move on.

A similar sentiment to Weil’s on working out an analogy:

Gone is the analogy: gone are the two theories, their conflicts and their delicious reciprocal reflections, their furtive caresses, their inexplicable quarrels; alas, all is just one theory, whose majestic beauty can no longer excite us. Nothing is more fecund than these slightly adulterous relationships; nothing gives greater pleasure to the connoisseur, whether he participates in it, or even if he is an historian contemplating it retrospectively, accompanied, nevertheless, by a touch of melancholy. The pleasure comes from the illusion and the far from clear meaning; once the illusion is dissipated, and knowledge obtained, one becomes indifferent at the same time

A host of balances to be struck then, including that between generating enough interest to bring in sufficiently many young people without creating a gold rush.

Posted by: David Corfield on November 24, 2006 10:10 AM | Permalink | Reply to this

Re: Klein 2-Geometry VII

When Street retires, I believe there will be no full professor at a research university in Australia who does category theory full time.

Don’t forget R.F.C. (“Bob”) Walters. These days, his and his students’ research interests tend strongly toward applications of category theory to CS, and he’s done quite a significant amount of work in pure category theory over the decades.

Posted by: Todd Trimble on November 24, 2006 2:26 AM | Permalink | Reply to this

Re: Klein 2-Geometry VII

Whoops – Walters is no longer listed as faculty at the University of Sydney; his home is now in Italy (I knew he’s had long-standing connections with the Italian categorical community). So AFAIK you’re right, John – Street is the only categorist who is a full professor at an Australian research university. Amazing.

Posted by: Todd Trimble on November 24, 2006 4:43 PM | Permalink | Reply to this

n-Quotients

Turning my eyes regretfully down from the gods on the heights of Olympus, I shall continue ploughing my lonely furrow at its base.

I have been thinking about quotients recently. I think I understand some things better now, so I shall put down my observations in case they are helpful to anybody else.

Weak, firm and strict quotients

We start with a groupoid. We want to think of this as a ‘weak quotient’. This usually means the result of acting on a set with a group. For each member of the group that sends an element aa of the set to an element bb, we add one morphism afb.a\overset{f}{\rightarrow}b. However, if, instead of considering an action (functor) from a group to Set, we consider more generally a functor from a groupoid to Set, sending all objects in the groupoid to the same set in Set, then we can think of any groupoid we like as being a weak quotient.

On the other hand, we have also the “strict” quotient, which is the set that we get by setting all isomorphic objects in the groupoid equal to one another and throwing away the morphisms (except identities).

I’d like to insert an intermediate level quotient. In this, we have neither a set of morphisms between two objects (as with the weak quotient), nor a “tautology” of morphisms (i.e. only the necessary identity morphisms) as with the strict quotient, but a truth value of morphisms. That is to say, if the weak quotient has at least one morphism between two objects, the intermediate quotient has exactly one. This is nothing other than a set together with an equivalence relation on it. This is nice, since an equivalence relation is traditionally seen as the intermediate step between a group action and a partition considered as a set.

What we are doing here is taking the obvious functor from the weak quotient to the strict quotient and factorising it into a full bit and a faithful bit.

Instead of this, we can try doing the steps in a different order. Instead of first reducing a set of morphisms to just one (per pair of objects), and then identifying objects, we can instead try identifying the objects first. To do this, we need to pick a particular isomorphism between each pair of objects, and “contract it down to the identity”, dragging the two objects together into one, and forcing the other isomorphisms and automorphisms to merge in a particular way. This gives us a set of groups. We can, if we wish, then force these down into a mere set of elements by identifying all automorphisms of a given object.

The fact that we can do this, merging at the 0-morphism level while failing to do so at the 1-morphism level, makes me wonder if perhaps we should have had a preliminary stage in which we add one 2-morphism from each 1-morphism to each other 1-morphism between the same objects, and then identified 1-morphisms by contraction of the 2-morphisms.

Quotients of groups

A group GG can act not only on an arbitrary set, but specifically on itself, by left multiplication. This gives the group a groupoid structure. The morphisms inherit a group operation from the group itself, and this turns the group into a 2-group GGG\rightarrow G, with the delta map being the identity, and the action being by conjugation.

We can also get a subgroup HH of GG to act on the whole group by multiplication. The resulting weak quotient is also a 2-group HGH\rightarrow G in the same way, that is with the delta map the inclusion and the action upwards by conjugation. More interestingly, the intermediate quotient is the equivalence relation which partitions the elements into the various cosets of the subgroup, while the strict quotient is precisely the quotient of the group by its subgroup.

We can also homomorphically map some other group KK into a subgroup HH of the group GG, and so long as the conjugation action of the whole group on the subgroup lifts correctly to the action of GG on K,K, we get a general strict 2-group like this. So a strict 2-group is a generalised weak quotient of groups.

Parallel morphisms and boundaries

Time for a parenthetical comment. When we are identifying kk-morphisms with each other for some k,k, there is a condition that they should have the same source and target k1k-1-morphisms: that they should be “parallel”. How does this work for objects (0-morphisms)? Presumably they should have the same “-1-morphisms” as each other. What does this mean? Well, we can see by inspection that the actual condition is that the objects should belong to the same component of the groupoid. That’s the only way there can be 1-morphisms between them that can be contracted to the identity.

This is interesting, because in a cell complex, if the 0-cells are the vertices, then the -1-cells are the components, and the boundary of a vertex is the component it belongs to.

More generally, the condition that two kk-morphisms should have the same source and target k1k-1-morphisms if they are to be identified, or, more generally, if they are to have any k+1k+1-morphisms between them, amounts to saying that they should have the same boundary, where their boundary is some sort of formal difference of their source and target. In terms of the k+1k+1-morphisms that putatively connect them, this amounts to saying that the boundaries of their source and target should be the same, i.e. have a difference of zero, or … OK, you’re way ahead of me … that the boundary of a boundary should be zero.

2-actions and 2-quotients

If a 1-action amounts to throwing in some morphisms to a set to form a groupoid, then a 2-action should amount to throwing in some 1-morphisms and 2-morphisms between them. Obviously, the 2-group (or 2-groupoid) that is the source of the action should send its own 1-morphisms to the new 1-morphisms and its 2-morphisms to the new 2-morphisms.

If the acted-upon object is a not merely a set, but a groupoid, we expect the 2-action to be by functors and natural isomorphisms between them. The result is, I think, an internal groupoid in the (or a) 2-category of groupoids.

More specifically, a 2-group HδGH\overset{\delta}{\rightarrow}G can act on itself. The action of GG is on itself is by multiplication, and its action on HH is the one given in the definition of the 2-group. In addition, the elements of Ker(δ)Ker(\delta) can act on HH by multiplication. The restriction to Ker(δ)Ker(\delta) is to ensure that the resulting 2-morphisms only send morphisms in HH to each other if they go between the same objects (i.e. if they differ by an automorphism of those objects, i.e. a member of Ker(δ)Ker(\delta).)

We can then repeat the same process as before, taking a subgroup of Ker(δ)Ker(\delta) or homomorphically mapping another group KK into it, to get a 3-group Kδ 1Hδ 0GK\overset{\delta_1}{\rightarrow}H\overset {\delta_0}{\rightarrow}G.

The condition that the image of δ 1\delta_1 lies in the kernel of δ 0\delta_0, or, in effect, that δ 2=0\delta^2=0 is of course precisely required by the condition that the boundary of a boundary be zero.

I have more things to say, but I’m not feeling too well just at the moment, so that’s all I can manage for now.

As always, apologies if I have been

a) too obvious, or
b) too obscure, or
c) both

Posted by: Tim Silverman on November 21, 2006 9:24 PM | Permalink | Reply to this

Re: n-Quotients

Weak, firm and strict quotients

On the other hand, we have also the “strict” quotient, which is the set that we get by setting all isomorphic objects in the groupoid equal to one another and throwing away the morphisms (except identities).

I like to think of this as identifying parallel morphisms. (This identification is what a functor from the weak quotient to the strict quotient does; it must take a value on all morphisms in the weak quotient.)

This is nothing other than a set together with an equivalence relation on it.

This is sometimes called a setoid. This term is (as far as I’ve seen) mostly used in type-theoretic approaches to the foundations of mathematics, but it’s an important idea in its own right.

In this, we have neither a set of morphisms between two objects (as with the weak quotient), nor a “tautology” of morphisms (i.e. only the necessary identity morphisms) as with the strict quotient, but a truth value of morphisms.

But in this case, I don’t think that this intermediate step (the ‘firm’ quotient?) is really appropriate. The reason is that, already back when it was a groupoid, we really shouldn’t think about equality of objects as a meaningful notion. So here, whether objects are strictly equal (identical) is unimportant, only their truth-value hom-set, which is also called their (meaningful) equality in the (strict) quotient set, matters.

In other words, a set already consists of a bunch of objects and truth-value hom-sets (called equality) between them; adding an additional equivalence relation (making it a setoid) consists in adding a second batch of truth-value hom-sets. Now, sometimes people like to consider groupoids up to ‘isomorphism’ (as that term is traditionally used between groupoids and categories), so their groupoids are indeed equipped with a batch of truth-value hom-sets (called equality) in addition to the set-valued hom-sets (whose elements are called morphisms). But normally, the truth-value hom-sets are quite irrelevant, and only the set-valued hom-sets matter; this is the view we get when we consider groupoids up to ‘equivalence’ (that being the traditional term used between groupoids and categories).

In other other words, the firm quotient is equivalent (as a groupoid) to the strict quotient; each may be seen (up to equivalence) as coming from the weak quotient by identifying parallel morphisms. Identifying isomorphic objects is irrelevant.

What we are doing here is taking the obvious functor from the weak quotient to the strict quotient and factorising it into a full bit and a faithful bit.

Unless I’m misunderstanding what you’re talking about (in which case my previous paragraphs must seem gibberish to you), then this is simply incorrect. (I mean that even if one disagrees with my philosophy in the previous paragraphs, still one should be able to agree with me about this fact.) The functor from the weak quotient to the strict quotient is already full (and surjective on objects as well), because every (necessarily identity) morphism in the strict quotient comes from a morphism in the weak quotient. It’s true that the functor from the intermediate quotient to the strict quotient is faithful, but it too is both full and surjective (thus an equivalence).

[…] pick a particular isomorphism between each pair of objects, and “contract it down to the identity”, dragging the two objects together into one, and forcing the other isomorphisms and automorphisms to merge in a particular way. This gives us a set of groups.

This set of groups (forming a groupoid as their disjoint union) is equivalent to the weak quotient. Indeed, it is simply a skeleton of the weak quotient. (In other words, every groupoid is equivalent to a set of groups; this explains how people can get away with resisting groupoids.)

Quotients of groups

This gives the group a groupoid structure. […] and this turns the group into a 2-group […].

This is an important way of getting groupoids. But it only gives a 2group if the group is abelian!

We can also homomorphically map some other group K into a subgroup H of the group G, and so long as the conjugation action of the whole group on the subgroup lifts correctly to the action of G on K, we get a general strict 2-group like this.

In this most general situation, we also need that K (hence H) be abelian (although G need not be).

These matters are discussed in HDA5.

Parallel morphisms and boundaries

Now we are leaving the realm of facts to return to matters of (at least potential) philosophical dispute.

When we are identifying […] objects (0-morphisms)? Presumably they should have the same “-1-morphisms” as each other. What does this mean? Well, we can see by inspection that the actual condition is that the objects should belong to the same component of the groupoid.

That is the condition for identifying that you apply when forming the strict quotient from the weak quotient, although I would argue that this identification is really vacuous (because it makes no difference, up to equivalence). In any case, this does not need to be the rule. That is, while one can (at least in a single step) identify only parallel morphisms, one may (when identifying objects) identify any two objects. (Indeed, this sort of thing happens when we pass from the original, unquotiented set to the strict quotient, or even the weak quotient if one considers adding isomorphism as just as good as identifying. Of course, the really interesting thing about the weak quotient, not appearing in the strict quotient, is that we also add additional automorphisms.)

Since all objects can be identified (even though in forming the strict quotient, they generally are not), all objects are parallel as 0morphisms. For a long time, this convinced me that every groupoid has exactly one (-1)morphism (which is why you don’t hear much about it), but there were still some niggles. Eventually (with Jim Dolan’s help) I realised that every occupied groupoid (that is a groupoid which at least one object) has a unique (-1)morphism, while the empty groupoid (that is a groupoid with no objects) has no (-1)morphism either.

So every groupoid has a truth value of (-1)objects and a set of objects (up to isomorphism —that is, a set of isomoprhism classes of objects— since equality of objects doesn’t respect equivalence of groupoids). If you want all morphisms, however, then you need a groupoid of morphisms. Or more generally, and ωgroupoid has an ngroupoid of (everything up to) nmorphisms (up to equivalence).

Posted by: Toby Bartels on November 22, 2006 11:05 PM | Permalink | Reply to this

Re: n-Quotients

Toby, many thanks for your comments. I’m not feeling well enough to say anything very intelligent in reply at the moment, but I will, I hope, when I am feeling better.

Posted by: Tim Silverman on November 23, 2006 10:37 PM | Permalink | Reply to this

Re: Klein 2-Geometry VII

On another thread I mention an introduction to stratified spaces by Paul Gunnells. He describes there what is the Schubert stratification of a Grassmanian, in particular G(4,2)G(4, 2). So you take a maximal flag, i.e., a sequence of subspaces of each dimension, each containing its predecessor. Then you see how R 2R^2 can intersect with this flag. It will do so in a non-decreasing sequence of natural numbers (0,a,b,c,2)(0, a, b, c, 2), where each entry is at most 1 more than its predecessor. Here (a,b,c)(a, b, c) can be (0, 0, 1), (0, 1, 1), (0, 1, 2), (1, 1, 1), (1, 1, 2), or (1, 2, 2). For each of these there is a Schubert cell of members of the Grassmanian, which fit together to make G(4,2)G(4, 2).

Now something we’d like to know is the structure of the Grassmanian of R a,bR^{a,b} sub vector 2-spaces of R p,qR^{p,q}. So presumably we need to choose a maximal flag, which in this case is not just a sequence of inclusions, but a lattice, with one entry for each pair of natural numbers no greater than (p,q)(p, q). Then we need to see how (a,b)(a, b) subspaces can intersect with this flag.

Even small dimensional examples produce large numbers of Schubert cells. Take the Grassmanian G((2,2),(1,1))G((2, 2), (1, 1)). A maximal flag has the shape of a diamond lattice with 9 entries. Intersection freedom is limited to 5 places, and in 4 of these only freedom in one of the dimensions. With a very quick calculation, there look to be 18 Schubert cells.

Posted by: David Corfield on November 25, 2006 11:39 AM | Permalink | Reply to this

Re: Klein 2-Geometry VII

Hmm. This can’t be right. There can’t be that much freedom for the growth of the intersection dimensions. As I noted back in September the 2-Grassmanian has the product of Grassmanians as objects. So here we should expect G(2,1)×G(2,1)G(2, 1) \times G(2, 1) worth of objects. So we might expect four Schubert cells. That looks right, and maybe not too interesting.

Posted by: David Corfield on November 25, 2006 8:16 PM | Permalink | Reply to this

Re: Klein 2-Geometry VII

Back in September David wrote:

…the 2-space of (n,0)(n, 0) sub-2-spaces of k p,qk^{p,q} has the Grassmanian G n,pG_{n,p} as objects and linear maps from k nk^n to k qk^q as automorphisms.

That sounds right to me.

While we’re at it, isn’t then the set of (a,b)(a,b) sub 2-spaces of k c,dk^{c,d} just G a,c×G b,dG_{a,c} \times G_{b,d} of objects each with linear maps k ak^a to k dk^d worth of automorphisms?

That sounds right. Let me talk it through. Once we’ve chosen skeletal representatives of our 2-vector spaces, we’re guessing an inclusion k a,bk c,d k^{a,b} \hookrightarrow k^{c,d} just amounts to an inclusion of objects k ak c k^a \hookrightarrow k^c together with an inclusion of morphisms k bk d k^b \hookrightarrow k^d Here of course we’re using some seat-of-the-pants concept of inclusion for 2-vector spaces, and disregarding all the careful work we once put into clarifying that issue!

Anyway…

If we write ‘Mono’ for inclusions, we thus have Mono(k a,b,k c,d)=Mono(k a,k c)×Mono(k b,k d) Mono(k^{a,b},k^{c,d}) = Mono(k^a,k^c) \times Mono(k^b,k^d) The point of this exercise in notation is that I’m guessing we should define the Grassmannian of (a,b)(a,b)-dimensional subspaces of k c,dk^{c,d} by Gr(k a,b,k c,d)=Mono(k a,b,k c,d)/Aut(k a,b) Gr(k^{a,b},k^{c,d}) = Mono(k^{a,b},k^{c,d})/Aut(k^{a,b}) where the automorphisms k a,bk a,b k^{a,b} \stackrel{\sim}{\to} k^{a,b} act on the inclusions k a,bk c,d k^{a,b} \hookrightarrow k^{c,d} by composition.

But, I think we have Aut(k a,b)=Aut(k a)×Aut(k b) Aut(k^{a,b}) = Aut(k^a) \times Aut(k^b) So, we get Gr(k a,b,k c,d)= Gr(k^{a,b},k^{c,d}) = Mono(k a,b,k c,d)/Aut(k a,b)= Mono(k^{a,b},k^{c,d})/Aut(k^{a,b}) = [Mono(k a,k c)×Mono(k b,k d)]/[Aut(k a)×Aut(k b)]= [Mono(k^a,k^c) \times Mono(k^b,k^d)]/[Aut(k^a) \times Aut(k^b)] = Mono(k a,k c)/Aut(k a)  ×  Mono(k b,k d)/Aut(k d)= Mono(k^a,k^c)/ Aut(k^a)     \times     Mono(k^b,k^d)/ Aut(k^d) = Gr(k a,k c)×Gr(k b,k d) Gr(k^a,k^c) \times Gr(k^b, k^d)

So, yeah - these categorified Grassmannians seem to have, as their space of objects, a product of two Grassmannians.

And, I think each of these objects has automorphisms as you describe, too.

So, I don’t think the theory of Schubert cells gets vastly more difficult - or more interesting.

Posted by: John Baez on November 26, 2006 8:21 AM | Permalink | Reply to this
Read the post Klein 2-Geometry VIII
Weblog: The n-Category Café
Excerpt: Continuing Klein 2-geometry
Tracked: December 1, 2006 9:52 AM
Read the post A Small Observation
Weblog: The n-Category Café
Excerpt: A duality in 2-vector spaces
Tracked: July 3, 2008 4:39 PM
Read the post A Small Observation
Weblog: The n-Category Café
Excerpt: A duality in 2-vector spaces
Tracked: July 3, 2008 4:42 PM

Post a New Comment